Mitochondrial-related genes PDK2, CHDH, and ALDH5A1 served as a diagnostic signature and correlated with immune cell infiltration in ulcerative colitis

Abstract

We conducted an investigation to determine the potential of mitochondrial-related genes as diagnostic biomarkers in ulcerative colitis (UC), while also examining their association with immune cell infiltration. To achieve this, we acquired four datasets pertaining to UC, which included gene expression arrays and clinical data, from the GEO database. Subsequently, we selected three signature genes (PDK2, CHDH, and ALDH5A1) to construct a diagnostic model for UC. The nomogram and ROC curves exhibited exceptional diagnostic efficacy. Following this, quantitative real-time polymerase chain reaction and western blotting assays validated the decreased mRNA and protein expression of PDK2, CHDH, and ALDH5A1 in the model of UC cells and dextran sulfate sodium salt (DSS)-induced mice colitis tissues, aligning with the findings in the risk model. This investigation suggested a negative correlation between the expression of ALDH5A1, CHDH, and PDK2 and the infiltration of M1 macrophages. Then, immunofluorescence analysis confirmed the augmented expression of CD86 in the tissue of mice subjected to DSS, while a diminished expression of ALDH5A1, CHDH, and PDK2 was observed. Consequently, it can be inferred that targeting mitochondria-associated genes, namely PDK2, CHDH, and ALDH5A1, holds potential as a viable strategy for prognostic prediction and the implementation of immune therapy for UC.

Keywords: aldehyde dehydrogenase 5 family member A1; choline dehydrogenase; immune infiltration; mitochondria; pyruvate dehydrogenase kinase 2; ulcerative colitis.

Figure 1

Construction of WGCNA co–expression network in training set. (A) The investigation involved the utilization of scale-free fit index and mean connectivity analysis across various soft-thresholding powers. (B) To identify and establish connections between related modules, the clustered dendrograms were truncated at 0.25 heights. (C) The WGCNA modules were assigned and represented by a colored horizontal bar in the cluster dendrograms. (D) The module feature genes were analyzed for collinearity and presented in a heatmap, where red indicates high correlation and blue indicates low correlation. (E) The topological overlap matrix (TOM) for each module was visualized in a WGCNA network heatmap. (F) Module-trait correlations. Positive correlations are represented by red and negative correlations by blue.

Figure 2

Construction of WGCNA co–expression network in validation set. (A) Scale-free fit index and mean connectivity analysis for various soft-thresholding powers. (B) Clustered dendrograms were cut at 0.25 heights to recognize and connect related modules. (C) Cluster dendrograms and module assignment for WGCNA modules. The colored horizontal bar represents the modules. (D) Collinear heat map of module feature genes. The color red represents a high correlation, whereas the color blue suggests the opposite results. (E) Visualization of the WGCNA network heatmap. The heatmap shows the topological overlap matrix (TOM) for each module. (F) Module-trait correlations. Positive correlations are represented by red and negative correlations by blue.

Figure 3

Identification of DEGs. (A) and (B) show the Volcano plot for DEGs between healthy controls and UC in the training and test groups, respectively. (C) and (D) Venn diagrams for intersections of DEGs and the WGCNA module in the training and test groups.

Figure 4

(A) Functional enrichment analysis. (B) and (C) GO and KEGG analysis of DEGs.

Figure 5

(A) Adjustment of feature selection in the minimum absolute shrinkage and selection operator model (lasso). (B) Random forest error rate versus the number of classification trees. (C) The top 12 relatively important genes. (D) Three algorithmic Venn diagram screening genes.

Figure 6

Construction and validation of the UC diagnostic column line graph model. (A, B) Column line graphs are utilized to predict the occurrence of UC in the training and validation cohorts. (C, D) ROC curves were created, with an area under the curve of 84.1305 in the training set and 72.8742 in the validation set.

Figure 7

Analysis of immune infiltration. (A) Bar plot bar plot depicting the 22 subpopulations of immune cells in UC and healthy control (HC) samples. (B) Correlation heatmap of immune cells in UC samples. (C) Violin plot illustrating the varying proportions of immune cells between UC and normal control (NC). (D–F) Bubble plots demonstrate the relationship between immune cells and specific signature genes (PDK2, CHDH, and ALDH5A1).

Figure 8

(A–C) The mRNA expression levels of the feature genes in the training cohort, as well as their ROC curves. (D–F) expression levels and ROC curves for the feature genes present in the training cohort. (G–I) the mRNA expression of three feature genes were validated using the LPS-induced NCM460 cells.

Figure 9

Expression of PDK2, CHDH, and ALDH5A1 in LPS-induced HIEC-6 cells and DSS-induced colitis mice. The relative mRNA (A) and protein (B) expression levels of PDK2, CHDH, and ALDH5A1 were conducted using the LPS-induced HIEC-6 cells. (C) Photographs were utilized to assess the variations in colon lengths among the groups. (D) The length of the colon in different groups. (E, F) Hematoxylin and eosin staining images and colonic histological scores of colon cross-sections in mice (n = 5). Scale bar = 50 μm. (G) Expression of PDK2, CHDH, and ALDH5A1 in the colon tissues of mice was detected using Western blotting. (H) Double immunofluorescence staining of PDK2 (labeled in green), CHDH (labeled in green), and ALDH5A1 (labeled in green) with M1 macrophages (CD86, labeled in red) at DSS induced mice. The nuclei were stained by DAPI (blue). n = 5 per group. Scale bar = 50 μm. Values are expressed as the mean ± sd (n = 5). ^*P < 0.05, ^**P < 0.01, ^***P < 0.001. Abbreviation: DSS: dextran sodium sulfate.